CS490D: Introduction to Data Mining Prof. Chris Clifton. Data Mining in Text

Transcription

1 CS490D: Introduction to Data Mining Prof. Chris Clifton March 26, 2004 Text Mining Data Mining in Text Association search in text corpuses provides suggestive information Groups of related entities Clusters that identify topics Flexibility is crucial Describe what an interesting pattern would look like What causes items to be considered associated same document, sequential associations,? Choice of techniques to rank the results Integrate with Information Retrieval systems Common base preprocessing (e.g. Natural Language processing) Need IR system to explore/understand text mining results 1

2 Why Text is Hard Lack of structure Hard to preselect only data relevant to questions asked Lots of irrelevant data (words that don t correspond to interesting concepts) Errors in information Misleading/wrong information in text Synonyms/homonyms: concept identification hard Difficult to parse meaning I believe X is a key player vs. I doubt X is a key player Sheer volume of patterns Need ability to focus on user needs Consequence for results: False associations Vague, dull associations What About Existing Products? Data Mining Tools Designed for particular types of analysis on structured data Structure of data helps define known relationship Small, inflexible set of pattern templates Text is free flow of ideas, tough to capture precise meaning Many patterns exist that aren t relevant to problem Experiments with COTS products on tagged text corpuses demonstrate these problems Discovery overload : many irrelevant patterns, density of actionable items too low Lack of integration with Information Retrieval systems makes further exploration/understanding of results difficult 2

3 What About Existing Products? Text Mining Information Retrieval Tools Text Mining is (mis?)used to mean information retrieval IBM TextMiner (now called IBM Text Search Engine ) DataSet These are Information Retrieval products Goal is get the right document May use data mining technology (clustering, association) Used to improve retrieval, not discover associations among concepts No capability to discover patterns among concepts in the documents. May incorporate technologies such as concept extraction that ease integration with a Knowledge Discovery in Text system What About Existing Products? Concept Visualization Goal: Visualize concepts in a corpus SemioMap SPIRE rch/spire.html Aptex Convectis High-level concept visualization Good for major trends, patterns Find concepts related to a particular query Helps find patterns if you know some of the instances of the pattern Hard to visualize rare event patterns 3

4 What About Existing Products? Corpus-Specific Text Mining Some Knowledge Discovery in Text products Technology Watch (patent office) hwatch.htm TextSmart (survey responses) Provide limited types of analyses Fixed questions to be answered Primarily high-level (similar to concept visualization) Domain-specific Designed for specific corpus and task Substantial development to extend to new domain or corpus What About Existing Products? Text Mining Tools True Text Mining just beginning to come to market Associations: ClearForest Semantic Networks: Megaputer s TextAnalyst IBM Intelligent Miner for Text (toolkit) Currently limited capabilities (but improving) Further research needed Directed research will ensure the right problems are solved Major Problem: Flood of Information Analyzing results as bad as reading the documents 4

5 Scenario: Find Active Leaders in a Region Goal: Identify people to negotiate with prior to relief effort Want general picture" of a region No expert that already knows the situation is available Problems: No clear central authority ; problems are regional Many claim power/control, few have it for long Must include all key players in a region Solution: Find key players over time Who is key today? Past players (may make a comeback) Example: Association Rules in News Stories Goal: Find related (competing or cooperating) players in regions Simple association rules (any associated concepts) gives too many results Flexible search for associations allows us to specify what we want: Gives fewer, more appropriate results Person1 Person2 Support Natalie Allen Linden Soles 117 Leon Harris Joie Chen 53 Ron Goldman Nicole Simpson Mobotu Sese Seko Laurent Kabila 10 Person1 Person2 Place Support Mobuto Laurent Kinshasa 7 Sese Seko Kabila 5

6 Conventional Data Mining System Architecture Data Repository Data Mining Tool Patterns Using Conventional Tools: Text Mining System Architecture Text Corpus Concept Extraction Repository Goal: Find Cooperating/ Combating Leaders in a territory Association Rule Product Person1 Person2 Natalie Allen Linden Soles 117 Leon Harris Joie Chen 53 Ron Goldman Nicole Simpson Mobotu Sese Seko Laurent Kabila 10 Too Many Results 6

7 Flexible Text Mining System Architecture Text Corpus Concept Extraction Repository Pattern Description Query P e r s o n 1 P e r s o n 2 P la c e Person1 and Person2 at Place Flocks DBMS M o b u to L a u re n t S e s e S e k o K a b ila K in s h a s a 7 Flexible: Adapts to new tasks Text Mining System Architecture Text Corpus Concept Extraction Repository Pattern Description Person1 then Person2 at Place in 1 month Query Flocks DBMS Person1 Person2 Place Avg. Days Mobuto Sese Seko Laurent Kabila Kinshasa 12 7

8 Example of Flexible Association Search Broadcast News Navigator Concept Correlation Tool News Source (default all) Broadcast Dates 60 Minutes ABC News Nightline ABC World News British Foreign CBS Evening News CNN Early Edition CNN Early Prime CNN Headline News All Dates From Date To Date Last Week Last Month Last Year 01 JAN JAN 1996 Find correlations between Person Location Organization, Person Location Organization, and Person Location Organization, reported in connection with the same location, within 2 days, having a minimum of 10 co-occurrences. Rank results by deviation from expected value. Text Databases and IR Text databases (document databases) Large collections of documents from various sources: news articles, research papers, books, digital libraries, messages, and Web pages, library database, etc. Data stored is usually semi-structured Traditional information retrieval techniques become inadequate for the increasingly vast amounts of text data Information retrieval A field developed in parallel with database systems Information is organized into (a large number of) documents Information retrieval problem: locating relevant documents based on user input, such as keywords or example documents 8

9 Information Retrieval Typical IR systems Online library catalogs Online document management systems Information retrieval vs. database systems Some DB problems are not present in IR, e.g., update, transaction management, complex objects Some IR problems are not addressed well in DBMS, e.g., unstructured documents, approximate search using keywords and relevance Basic Measures for Text Retrieval Relevant Relevant & Retrieved Retrieved All Documents Precision: the percentage of retrieved documents that are in fact relevant to the query (i.e., correct responses) { Relevant} { Retrieved} precision = { retrieved} Recall: the percentage of documents that are relevant to the query and were, in fact, retrieved { Relevant} { Retrieved} recall = { relevant} 9

10 Basic Concepts Information Retrieval Techniques(1) A document can be described by a set of representative keywords called index terms. Different index terms have varying relevance when used to describe document contents. This effect is captured through the assignment of numerical weights to each index term of a document. (e.g.: frequency, tf-idf) DBMS Analogy Index Terms Attributes Weights Attribute Values Information Retrieval Techniques(2) Index Terms (Attribute) Selection: Stop list Word stem Index terms weighting methods Terms Documents Frequency Matrices Information Retrieval Models: Boolean Model Vector Model Probabilistic Model 10

11 Boolean Model Consider that index terms are either present or absent in a document As a result, the index term weights are assumed to be all binaries A query is composed of index terms linked by three connectives: not, and, and or e.g.: car and repair, plane or airplane The Boolean model predicts that each document is either relevant or non-relevant based on the match of a document to the query Boolean Model: Keyword- Based Retrieval A document is represented by a string, which can be identified by a set of keywords Queries may use expressions of keywords E.g., car and repair shop, tea or coffee, DBMS but not Oracle Queries and retrieval should consider synonyms, e.g., repair and maintenance Major difficulties of the model Synonymy: A keyword T does not appear anywhere in the document, even though the document is closely related to T, e.g., data mining Polysemy: The same keyword may mean different things in different contexts, e.g., mining 11

12 Similarity-Based Retrieval in Text Databases Finds similar documents based on a set of common keywords Answer should be based on the degree of relevance based on the nearness of the keywords, relative frequency of the keywords, etc. Basic techniques Stop list Set of words that are deemed irrelevant, even though they may appear frequently E.g., a, the, of, for, to, with, etc. Stop lists may vary when document set varies Similarity-Based Retrieval in Text Databases (2) Word stem Several words are small syntactic variants of each other since they share a common word stem E.g., drug, drugs, drugged A term frequency table Each entry frequent_table(i, j) = # of occurrences of the word t i in document d i Usually, the ratio instead of the absolute number of occurrences is used Similarity metrics: measure the closeness of a document to a query (a set of keywords) Relative term occurrences v Cosine distance: 1 v2 sim( v1, v2) = v1 v2 12

13 CS490D: Introduction to Data Mining Prof. Chris Clifton April 2, 2004 Text Mining Indexing Techniques Inverted index Maintains two hash- or B+-tree indexed tables: document_table: a set of document records <doc_id, postings_list> term_table: a set of term records, <term, postings_list> Answer query: Find all docs associated with one or a set of terms + easy to implement do not handle well synonymy and polysemy, and posting lists could be too long (storage could be very large) Signature file Associate a signature with each document A signature is a representation of an ordered list of terms that describe the document Order is obtained by frequency analysis, stemming and stop lists 13

14 Vector Model Documents and user queries are represented as m-dimensional vectors, where m is the total number of index terms in the document collection. The degree of similarity of the document d with regard to the query q is calculated as the correlation between the vectors that represent them, using measures such as the Euclidian distance or the cosine of the angle between these two vectors. Latent Semantic Indexing (1) Basic idea Similar documents have similar word frequencies Difficulty: the size of the term frequency matrix is very large Use a singular value decomposition (SVD) techniques to reduce the size of frequency table Retain the K most significant rows of the frequency table Method Create a term x document weighted frequency matrix A SVD construction: A = U * S * V Define K and obtain U k,, S k, and V k. Create query vector q. Project q into the term-document space: Dq = q * U k * S k -1 Calculate similarities: cos α = Dq. D / Dq * D 14

15 Latent Semantic Indexing (2) Weighted Frequency Matrix Query Terms: - Insulation - Joint Probabilistic Model Basic assumption: Given a user query, there is a set of documents which contains exactly the relevant documents and no other (ideal answer set) Querying process as a process of specifying the properties of an ideal answer set. Since these properties are not known at query time, an initial guess is made This initial guess allows the generation of a preliminary probabilistic description of the ideal answer set which is used to retrieve the first set of documents An interaction with the user is then initiated with the purpose of improving the probabilistic description of the answer set 15

16 Types of Text Data Mining Keyword-based association analysis Automatic document classification Similarity detection Cluster documents by a common author Cluster documents containing information from a common source Link analysis: unusual correlation between entities Sequence analysis: predicting a recurring event Anomaly detection: find information that violates usual patterns Hypertext analysis Patterns in anchors/links Anchor text correlations with linked objects Motivation Keyword-Based Association Analysis Collect sets of keywords or terms that occur frequently together and then find the association or correlation relationships among them Association Analysis Process Preprocess the text data by parsing, stemming, removing stop words, etc. Evoke association mining algorithms Consider each document as a transaction View a set of keywords in the document as a set of items in the transaction Term level association mining No need for human effort in tagging documents The number of meaningless results and the execution time is greatly reduced 16

17 Text Classification(1) Motivation Automatic classification for the large number of on-line text documents (Web pages, s, corporate intranets, etc.) Classification Process Data preprocessing Definition of training set and test sets Creation of the classification model using the selected classification algorithm Classification model validation Classification of new/unknown text documents Text document classification differs from the classification of relational data Document databases are not structured according to attributevalue pairs Text Classification(2) Classification Algorithms: Support Vector Machines K-Nearest Neighbors Naïve Bayes Neural Networks Decision Trees Association rule-based Boosting 17

18 Document Clustering Motivation Automatically group related documents based on their contents No predetermined training sets or taxonomies Generate a taxonomy at runtime Clustering Process Data preprocessing: remove stop words, stem, feature extraction, lexical analysis, etc. Hierarchical clustering: compute similarities applying clustering algorithms. Model-Based clustering (Neural Network Approach): clusters are represented by exemplars. (e.g.: SOM) Goal: Automatically Identify Recurring Topics in a News Corpus Started with a user problem: Geographic analysis of news Idea: Segment news into ongoing topics/stories How do we do this? What we need: Topics Mnemonic for describing/remembering the topic Mapping from news articles to topics Other goals: Gain insight into collection that couldn t be had from skimming a few documents Identify key players in a story/topic 18

19 User Problem: Geographic News Analysis TopCat identified separate topics for U.S. embassy bombing and counterstrike. Bombing List of Counter- Topics strike A Data Mining Based Solution Idea in Brief A topic often contains a number of recurring players/concepts Identified highly correlated named entities (frequent itemsets) Can easily tie these back to the source documents But there were too many to be useful Frequent itemsets often overlap Used this to cluster the correlated entities But the link to the source documents is no longer clear Used topic (list of entities) as a query to find relevant documents to compare with known mappings Evaluated against manually-categorized ground truth set Six months of print, video, and radio news: 65,583 stories 100 topics manually identified (covering 6941 documents) 19

20 TopCat Process Identify named entities (person, location, organization) in text Alembic natural language processing system Find highly correlated named entities (entities that occur together with unusual frequency) Query Flocks association rule mining technique Results filtered based on strength of correlation and number of appearances Cluster similar associations Hypergraph clustering based on hmetis graph partitioning algorithm (based on (Han et. al. 1997)) Groups entities that may not appear together in a single broadcast, but are still closely related Preprocessing Identify named entities (person, location, organization) in text Alembic Natural Language Processing system Data Cleansing: Coreference Resolution Used intra-document coreference from NLP system Heuristic to choose global best name from different choices in a document Eliminate composite stories Heuristic - same headline monthly or more often High Support Cutoff (5%) Eliminate overly frequent named entities (only provide common knowledge topics) 20

21 Named Entities vs. Full Text Corpus contained about 65,000 documents. Full text resulted in almost 5 million unique worddocument pairs vs. about 740,000 for named entities. Prototype was unable to generate frequent itemsets at support thresholds lower than 2% for full text. At 2% support, one week of full text data took 30 times longer to process than the named entities at 0.05% support. For one week: 91 topics were generated with the full text, most of which aren t readily identifiable. 33 topics were generated with the named-entities. Frequent Itemsets Israel State West Bank Netanyahu Albright Arafat Iraq State Albright 479 Israel Jerusalem West Bank Netanyahu Arafat Gaza Netanyahu 39 Ramallah Authority West Bank Iraq Israel U.N. 39 Query Flocks association rule mining technique frequent itemsets with 0.05% support Results filtered based on strength of correlation and support Cuts to 3129 frequent itemsets Ignored subsets when superset with higher correlation found 449 total itemsets, at most 12 items (most 2-4) 21

22 Cluster similar associations Clustering Hypergraph clustering based on hmetis graph partitioning algorithm (adapted from (Han et. al. 1997)) Groups entities that may not appear together in a single broadcast, but are still closely related Ramallah Gaza Authority Arafat West Bank Netanyahu { v P} { v e} { v e} n j= 1 n i= 1 Weight( cut_edges ) Weight( original_edges Jerusalem U.N. i Israel j ) Albright Iraq State Mapping to Documents Mapping Documents to Frequent Itemsets easy Itemset with support k has exactly k documents containing all of the items in the set. Topic clusters harder Topic may contain partial itemsets Solution: Information Retrieval Treat items as keys to search for Use Term Frequency/Inter Document Frequency as distance metric between document and topic Multiple ways to interpret ranking Cutoff: Document matches a topic if distance within threshold Best match: Document only matches closest topic 22

23 Merging Topics still to fine-grained for TDT Adjusting clustering parameters didn t help Problem was sub-topics Solution: Overlap in documents Documents often matched multiple topics Used this to further identify related topics Marriage Parent/Child i TFIDF TFIDF N documents ia i documents i documents ia TFIDF N ib TFIDF ib N i documents TFIDF TFIDF i documents ip TFIDF ic N ic N TopCat: Examples from Broadcast News LOCATION Baghdad PERSON Saddam Hussein PERSON Kofi Annan ORGANIZATION United Nations PERSON Annan ORGANIZATION Security Council LOCATION Iraq LOCATION Israel PERSON Yasser Arafat PERSON Walter Rodgers PERSON Netanyahu LOCATION Jerusalem LOCATION West Bank PERSON Arafat 23

24 TopCat Evaluation Tested on Topic Detection and Tracking Corpus Six months of print, video, and radio news sources 65,583 documents 100 topics manually identified (covering 6941 documents) Evaluation results (on evaluation corpus, last two months) Identified over 80% of human-defined topics Detected 83% of stories within human-defined topics Misclassified 0.2% of stories Results comparable to official Topic Detection and Tracking participants Slightly different problem - retrospective detection Provides mnemonic for topic (TDT participants only produce list of documents) Project Participants MITRE Corporation Modeling intelligence text analysis problems Integration with information retrieval systems Technology transfer to Intelligence Community through existing MITRE contracts with potential developers/first users Stanford University Computational issues Integration with database/data mining Technology transfer to vendors collaborating with Stanford on other data mining work Visitors: Robert Cooley (University of Minnesota, Summer 1998) Jason Rennie (MIT, Summer 1999) 24

25 MITRE internal: Where we re going now: Use of the Prototype Broadcast News Navigator GeoNODE External Use: Both Broadcast News Navigator and GeoNODE planned for testing at various sites GeoNODE working with NIMA as test site Incorporation in DARPA-sponsored TIDES Portal for Strong Angel/RIMPAC exercise this summer Exercise Strong Angel June 2000 Hawaii The scenario Humanitarian Assistance Increasing violence against Green minority in Orange Green minority refugees massing in border mountains Ethnic Green crossing into Green, though Orange citizens Live bomblets found near roads Basics in short supply water, shelter, medical care 25

26 What We ve Learned: Recommendations/Thoughts for Further Work Want flexibility in describing patterns What lends support to an association (e.g. across hyperlink; combining sequential, standard associations) Type of associated entity important in describing pattern Major risk: density of good stuff in results too low Problem isn t wrong results, but uninteresting results Simple support/confidence rarely appropriate for text Support a range of metrics - no single proper measure Cleaning and Mining as part of same process Human cost of pre-mining cleansing too high Human feedback on mining results (may alter results) What we see in the Future: COTS support for Data Mining in Text Working with vendors to incorporate query flocks technology in DBMS systems Stanford University working with IBM Almaden Research Working with vendors to incorporate text mining in information retrieval systems MITRE discussing technology transition with Manning&Napier Information Services, Cartia More Research needed What are the types of analyses that should be supported? What are the right relevance measures to find interesting patterns, and how do we optimize these? What additional capabilities are needed from concept extraction? 26

27 Potential Applications Topic Identification Identify by different types of entities (person / organization / location / event /?) Hierarchically organize topics (in progress) Support for link analysis on Text Tools exist for visualizing / analyzing links (e.g. NetMap) Text mining detects links -- giving link analysis tools something to work with Support for Natural Language Processing / Document Understanding Synonym recognition -- A and B may not appear together, but they each appear with X, Y, and Z -- A and B may be synonyms Prediction: Sequence analysis (in progress) Similarity Search in Multimedia Data Description-based retrieval systems Build indices and perform object retrieval based on image descriptions, such as keywords, captions, size, and time of creation Labor-intensive if performed manually Results are typically of poor quality if automated Content-based retrieval systems Support retrieval based on the image content, such as color histogram, texture, shape, objects, and wavelet transforms 27

28 Queries in Content-Based Retrieval Systems Image sample-based queries Find all of the images that are similar to the given image sample Compare the feature vector (signature) extracted from the sample with the feature vectors of images that have already been extracted and indexed in the image database Image feature specification queries Specify or sketch image features like color, texture, or shape, which are translated into a feature vector Match the feature vector with the feature vectors of the images in the database 28